AI hiring statistics in 2026: the numbers that survive scrutiny

The AI hiring statistics worth citing start with SHRM. In its most recent measurement, fielded December 5 to 23, 2025, 39% of organizations use AI somewhere in HR, and recruiting is the single most common place they use it, at 27%. That is SHRM's State of AI in HR 2026 report, published March 31, 2026, which SHRM describes as drawing on 1,908 HR professionals.

Most published AI hiring statistics run much higher, because they count intent, pilots, and any single AI-assisted task.

That gap is what this page is about. I work on growth at Expert Hire. We run live conversational AI interviews with live coding and produce an explainable report card. So we sell into this category, and we have every incentive to quote the big numbers.

We are not going to. Every figure below carries the number, the publisher, the sample size, the field date, and a link to the original.

Search for AI recruiting statistics, or AI in hiring statistics 2026, or AI recruitment adoption statistics, and you will get wildly different answers to the same question. The spread is not noise. It is what gets counted.

Key Takeaways

Our position: the widely repeated AI hiring adoption numbers are inflated, and the honest figure for AI in recruiting is around a quarter to a third of organizations, not most of them.
SHRM's own two surveys show why numbers diverge. 26% of all organizations used AI in HR in 2024, while 64% of AI-adopting organizations used it for recruiting. Same survey, different denominator.
Public opinion is worse than vendors admit. 66% of Americans told Pew they would not want to apply to an employer using AI in hiring decisions.
Measured bias is real and peer-reviewed. Three production language models preferred white-associated names 85% of the time against 9% for Black-associated names.
Of 391 US employers checked against NYC Local Law 144, 18 had posted a bias audit. The researchers stress this cannot be read as a non-compliance rate, because employers decide for themselves whether a tool is in scope.
Adoption is not effectiveness. 56% of HR functions do not formally measure whether their AI investment worked.

AI hiring statistics at a glance: the numbers worth citing

Each figure below is traceable to a primary publisher with a stated sample. Anything that resolved only to an aggregator quoting another aggregator was cut rather than softened, which is why this list is shorter than most.

39% of organizations currently use AI in their HR functions, and another 23% use AI elsewhere in the business, per SHRM's State of AI in HR 2026 report, a survey of 1,908 HR professionals fielded December 5 to 23, 2025 via SHRM's Voice of Work Research Panel and published March 31, 2026.
Recruiting is the most common HR application of AI, at 27%, ahead of HR technology at 21% and learning and development at 17% (SHRM, State of AI in HR 2026, fielded December 2025).
60% of organizations with 5,000 or more employees use AI in HR, against 39% across all organization sizes (SHRM, State of AI in HR 2026, fielded December 2025).
26% of organizations used AI to support any HR-related activity in SHRM's 2024 Talent Trends AI findings, a survey of 2,366 US HR professionals fielded January 10-19, 2024.
Among the organizations in that 2024 survey that had adopted AI for HR, 64% used it for recruitment, interviewing, and hiring, the single largest use case (SHRM, 2024 Talent Trends).
56% of HR functions do not formally measure the success of their AI investments, and only 16% use ROI as a metric (SHRM, State of AI in HR 2026, fielded December 2025).
66% of Americans say they would not want to apply for a job with an employer that uses AI to help make hiring decisions, against 32% who would, per the Pew Research Center, surveying 11,004 US adults, fielded December 12-18, 2022 and published April 20, 2023.
Americans oppose AI making final hiring decisions by roughly ten to one, and 41% oppose AI being used to review job applications (same Pew survey).
Three production language models ranking resumes against more than 500 real job listings preferred white-associated names 85% of the time against 9% for Black-associated names, and male-associated names 52% against 11% for female-associated names, across more than three million comparisons (Wilson and Caliskan, AIES 2024, open-access preprint).
Of 391 US employers checked by 155 student investigators for NYC Local Law 144 compliance, 18 had posted a bias audit report and 13 had posted a transparency notice (Wright et al., ACM FAccT 2024). The authors warn this is not a non-compliance rate, because employers decide for themselves whether a tool is in scope.

How many companies use AI in hiring

Somewhere between a quarter and two fifths, depending on the year and the exact question. SHRM's most recent figure, fielded in December 2025, is 39% of organizations using AI somewhere in HR. Its January 2024 survey of 2,366 HR professionals put that same measure at 26%. Both numbers are SHRM's, both are honestly reported, and the jump reflects real growth over two years.

For recruiting specifically, SHRM puts it at 27% of organizations, the largest single HR application of AI, ahead of HR technology at 21% and learning and development at 17%.

Set that against the 39% using AI anywhere in HR, and roughly two thirds of AI-adopting organizations are using it in recruiting. That tracks SHRM's 2024 finding that 64% of AI adopters used it for recruitment, interviewing, and hiring, which is about as much corroboration as this category offers.

Where in the funnel the AI sits is the part almost nobody reports. SHRM's 2024 survey asked where that AI actually sits. Among organizations using it for recruiting, interviewing, or hiring, 65% used it to write job descriptions and 42% to target job postings. Another 34% used it to screen resumes, and 22% to schedule interviews.

Only 13% used it to pre-select applicants for interviews, 7% to conduct pre-screening interviews, and 3% to analyze interview performance. The decision itself stays overwhelmingly human, which our guide to how AI is changing hiring covers in narrative form.

Why AI hiring statistics disagree with each other

Two honestly reported surveys can produce 27% and 88% for what sounds like the same question. Here is the taxonomy of why, and you can apply it to any statistic you find.

The denominator moves. SHRM's 2024 survey found 64% of AI-adopting organizations using AI for recruiting, and 26% of all organizations using AI in HR at all. Multiply those and you get roughly 17% of all organizations, not 64%. Aggregators quote the 64%.
"AI in HR" is not "AI in recruiting." Payroll anomaly detection and an AI chatbot answering benefits questions both count as AI in HR. Neither touches hiring.
Any AI touch is not AI in the decision. A recruiter using an assistant to rewrite a job description counts as adoption in most survey instruments. SHRM's 2024 breakdown sizes the gap: 65% of AI-using recruiting teams used it to write job descriptions, 7% to run a pre-screening interview.
A vendor customer panel is not a random sample. If you survey the people who bought your product, adoption looks near universal. Self-selection inflates every vendor-run adoption figure in this category.
Self-report is not telemetry. SHRM reports 87% saying AI improved efficiency and 75% saying it improved work quality (State of AI in HR 2026, fielded December 2025). Those are assessments by the teams that chose and paid for the tools, not measured outcomes.

One more, smaller and instructive, and it sits inside our own headline source. SHRM's public pages describe State of AI in HR 2026 as drawing on 1,908 HR professionals. Write-ups quoting the report's methodology statement give 1,722 participants, fielded December 5 to 23, 2025. The full report is gated, so we cite the number SHRM publishes openly and flag the gap rather than pick a side.

What candidates think about being screened by AI

Badly, and the best available US data is old enough that you must say so. The Pew Research Center surveyed 11,004 US adults between December 12 and 18, 2022, and published on April 20, 2023. 66% said they would not want to apply for a job with an employer that uses AI to help make hiring decisions.

The same survey found Americans opposing AI making final hiring decisions by roughly ten to one, with 41% opposing AI reviewing applications at all. That last number is worth sitting with. Opposition to screening is much softer than opposition to deciding, which maps almost exactly onto where employers actually deploy AI.

We publish this knowing it cuts against us. Anyone selling AI interviewing who quotes candidate-experience statistics without quoting Pew's 66% is cherry-picking. The more useful framing is our comparison of AI interviews and human interviews, which is about signal quality rather than sentiment.

Bias and compliance: what the research actually measured

What the peer-reviewed bias research found

Kyra Wilson and Aylin Caliskan tested three production language models from Mistral AI, Salesforce, and Contextual AI, ranking resumes against more than 500 real job listings. Across more than three million comparisons, the models preferred white-associated names 85% of the time against 9% for Black-associated names, and male-associated names 52% against 11% for female-associated names.

The intersectional finding is the one people skip. The systems never preferred Black male-associated names over white male-associated names. The work was presented at the AAAI/ACM Conference on AI, Ethics, and Society in San Jose on October 22, 2024. The paper carries the full method, with an open-access preprint and a plain-language University of Washington summary.

This is a measurement of general-purpose language models used as resume rankers, not of every tool marketed as AI hiring. It remains the strongest evidence that unstructured AI screening reproduces name-based bias at scale. It is why our approach to reducing bias in hiring starts from a structured, role-tuned rubric that is set before the interview runs, rather than a similarity score.

The compliance gap between the rule and reality

NYC Local Law 144 requires an annual independent bias audit, published results, and candidate notice for automated employment decision tools, per the NYC Department of Consumer and Worker Protection. Public evidence of compliance is close to absent. Of 391 US employers checked by 155 student investigators, 18 had posted an audit report and 13 a transparency notice.

The researchers call this null compliance, and the distinction matters. Employers decide for themselves whether a tool is in scope, so an absent audit cannot be scored as non-compliance. The denominator of employers actually running such a tool is unknown, which is the paper's central point rather than a footnote to it. The sample also skews toward credentialed professional roles, because it was drawn from employers that hire Cornell graduates.

Of those, 267 had open NYC roles. The Cornell, Data and Society, and Consumer Reports team found 14 audit reports and 12 notices. Only 11 employers published both in a form the law would accept. Nearly every published audit reported an impact ratio above 0.8, the four-fifths threshold.

Federal direction moved the other way in 2025. The EEOC's May 2023 technical assistance on adverse impact in algorithmic selection was removed from eeoc.gov, and its canonical URL returned a 404 when we checked on September 22, 2026. The ACLU of Massachusetts preserved an archived copy of the withdrawn guidance, which is the only way left to read it.

Executive Order 14281, signed April 23, 2025, directs agencies to deprioritize disparate-impact enforcement. Title VII itself did not change.

Meanwhile the EU moved in the opposite direction. Annex III of Regulation (EU) 2024/1689, the EU AI Act, classifies AI used for recruitment, targeted job advertising, application filtering, and candidate evaluation as high-risk. We keep the jurisdiction-by-jurisdiction detail in our guide to AI hiring laws.

How to read an AI hiring statistic without getting burned

Four questions kill most of the numbers circulating in this category, and they take about thirty seconds each.

Who paid for the study? A vendor surveying its own customers is publishing a marketing asset. That does not make it false, it makes it unrepresentative.
Who was sampled, and how many? No stated sample size means no statistic. "HR leaders say" is not a population.
What counted as "using AI"? Drafting a job description and auto-rejecting applicants are both "using AI" in most instruments and are not remotely the same decision.
When was it fielded? Publication date is not field date. The best candidate-sentiment data in this article was collected in December 2022 and is still being quoted as current.

Apply this to us too. Expert Hire's own platform numbers are vendor numbers, and you should discount them the way you would a competitor's. That is why we document the scoring approach and the research behind it on our methodology page rather than ask anyone to trust a score.

The same test applies to skills-based hiring and structured interview software. The underlying research there is strong. The vendor claims layered on top of it often are not.

Frequently asked questions

What percentage of companies actually use AI in hiring in 2026?

Around a quarter to a third. SHRM's State of AI in HR 2026 report draws on 1,908 HR professionals and was fielded December 5 to 23, 2025. It puts AI use anywhere in HR at 39% of organizations, with recruiting the leading application at 27%.

Adoption skews large: 60% among organizations with 5,000 or more employees. Figures above 70% almost always count intent, pilots, or AI use anywhere in the business.

Why do AI hiring statistics from different sources disagree so much?

Usually the denominator. SHRM's 2024 survey reported 64% of AI-adopting organizations using AI for recruiting and 26% of all organizations using AI in HR, which works out to roughly 17% of all organizations. Aggregators quote the 64% without the base. The second most common cause is a vendor surveying its own customer list and calling it a market survey.

How do candidates feel about being screened or interviewed by AI?

Skeptical, especially about decisions. Pew Research Center asked 11,004 US adults. Two thirds, 66%, would not want to apply to an employer using AI to help make hiring decisions. Opposition to AI making the final call ran about ten to one.

Opposition to AI reviewing applications was much lower, at 41%. That survey was fielded in December 2022, so treat it as the best available baseline rather than a current reading.

Is AI hiring legal, and which laws require a bias audit?

It is legal, with conditions that vary by jurisdiction. NYC Local Law 144 requires an annual independent bias audit, published results, and candidate notice for automated employment decision tools. Illinois regulates AI in video interviews, and the EU AI Act classifies recruitment and candidate evaluation as high-risk under Annex III. Federal US guidance loosened in 2025, but Title VII obligations are unchanged.

How many job applications are now written or submitted with AI?

Nobody has published a defensible number. Every widely quoted figure we chased, including the popular applications-per-minute statistic, traced to press coverage or to a vendor report with no disclosed sample, not to a primary publisher. The honest answer is that application volume is clearly up and the magnitude is unmeasured in public. Treat any precise figure on this with suspicion until someone publishes a method.

The numbers worth defending

If you take one thing from this page, make it the denominator question. Most of the inflation in AI hiring statistics is not fabrication, it is a true number reported against a base that quietly shifted between the survey and the headline. Recruiting leads every other HR function in AI adoption and still sits near a quarter to a third of organizations, not most of them.

We will re-verify every figure here quarterly and stamp the date, because a statistics page that goes stale under our name is worse than no page at all.

Reading how many companies claim AI is not the same as judging whether it holds a technical bar. Our AI interview platform page walks through how a round is scored and what lands on the report card, with the transcript behind every score.

By TK, Growth at Expert Hire. Last updated September 22, 2026. Reviewed by Anand Suresh, CPO at Expert Hire.